J. theor. Biol. (1995) 173, 41-50 Why Pelagic Planktivores should be Unselective Feeders JARL GISKEt AND ANNE GRO VEA SALVANES University of Bergen, Department of Fisheries and Marine Biology, Hoyteknologisenteret, N-5020 Bergen and tlnstitute of Marine Research, Boks 1870 Nordnes, N-5024 Bergen, Norway (Received on 29 January 1994, Accepted in revised form on 18 July 1994) Diet width theory is a branch of optimal foraging theory, used to predict which fractions of the potential food encountered should be pursued. For pelagic fish, it is generally assumed that light is the dominant stimulus for both prey encounter rate and mortality risk. In order to achieve encounter rates allowing selective feeding, the pelagic predator exposes itself to enhanced predation risk for a prolonged time. The gain in growth obtained by diet selection may seldom outweigh the fitness cost of increased mortality risk. More generally, pelagic feeders will have a higher reproductive rate by searching the depth where feeding will be encounter-limited, and hence be opportunistic feeders. Literature reports of pelagic diet selection either fail to distinguish between the catchability of the prey in a gear and the encounter rate with its predator or neglects the vertical structure in pelagic prey distribution that may give differences in diets for unselective predators operating at different depths. The principal differences between the pelagic habitat and habitats where diet selection will be expected will include one or both of the following: (i) the continuous and steep local (i.e. vertical) gradients in mortality risk and (ii) the lack of local shelter where a newly ingested meal may safely be digested. 1. Introduction needed to pursue, handle and swallow a single item" (Schoener, 1987: 12). Time elapsed due to digestion Optimal diet width theory was put forward to find the constraints are thus not incorporated in H. number of prey types an optimal predator should There is no emphasis of spatial differences in prey pursue. Under the assumption that the optimal for- distributions in equation (l)--they are eventually agers maximize net energetic gain rate, prey types reflected in the 2is. A selective diet implies that the could be ranked according to their net energetic gain predator encounters potential prey organims which it (E/H; energy per handling time) obtained for each does not chase. The fundamental parameter is the prey type. A diet including m - 1 prey types should predator-prey encounter rate 2j. (Often diet selection not be expanded to also include the m-th ranked item is claimed on basis of prey abundance indices instead if the average gain rate from the m- 1 prey types of encounter rates. Only when one can ascertain that would be higher than the gain rate from the m prey the predator and the scientist or the sampling tool has types: equal prey encounter rates, can abundance indices m--I be used. This is seldom the case in an aquatic E.,/H., <<. ~ 2j(Ej - E,,,Hj/H.,) (I) environment.) j=l As eqn (I) was developed in the early stages of (Schoener, 1971; Charnov, 1976; Townsend & optimal foraging theory (c.f. Schoener, 1987), the Winfield, 1985) where 2/, Ej and Hj are encounter rate, influence of predation risk was not mentioned in these energy content and handling time of prey type j, diet selection models. Townsend & Winfield (1985) respectively. (All symbols used in all equations are concluded that doing so "would be a significant step explained in Table 1.) Handling time is "the time forward". Gilliam (1990) combined the optimal diet 0022-5193/95/050041 + 10 $08.00/0 41 (" 1995 Academic Press Limited 42 J. GISKE AND A. G. V. SALVANES width equation with the "static optimization" (Montgomery, 1987; Jumper & Baird, 1991; Janssen equation for optimal habitat choice under predation & Corcoran, 1993; Montgomery & Milton, 1993), in risk (Werner & Gilliam, 1984) and developed a model depths where prey encounter rate is sufficient for for the influence of mortality risk on diet selection. selective feeding, vision will be the dominant cue for His variables included mortality risk while searching prey perception (e.g. Vinyard & O'Brien, 1976; Munz and handling in addition to those in eqn (1). He linked & McFarland, 1977; Confer et al., 1978; Howick & mortality risk to prey type, so that Ej/Hj is the energy O'Brien, 1983; Nicol, 1989). The model does not gain and MjHj is the mortality risk while handling apply to surface predators. Neither does it apply to prey j. Turner & Mitteibach (1990) used optimal diet aquatic birds, nor rarely to marine mammals, as width to calculate growth in bluegill sunfish in differ- they depend on regular surface encounters and are ent habitats, and predicted the optimal habitat under probably forced to feed in depths where diet selection the assumption that they prefer the habitat where may pay. mortality risk per growth rate is minimal. In Section 2.1 we first outline life-history models to The investigations of Werner, Mittelbach and col- show that feeding rate can be linked to fitness either leagues (e.g. Werner et al., 1983a, b; Mittelbach, 1983, by a reduction in generation time or by an increase in 1988; Werner & Hall, 1988; Osenberg & Mittelbach, fecundity. This is used to illustrate static optimization 1989; Turner & Mittelbach, 1990) are typical for a trade-offs between feeding and mortality risk for situation where diet selection is expected and import- different life histories and age groups. The fitness-gain ant. They studied shallow-water freshwater fish living of feeding is highest when increased feeding may in or near the vegetation. In this environment, light lead to reduction in generation time (c.f. Werner & intensity and prey encounter rate is high, and the Gilliam, 1984: "minimize mortality risk per feeding feeding rate of the fish during daytime will be limited rate"). In Section 2.2 we show that environmental by digestion rate and not by encounter rate. Also, in variability in mortality risk will be proportional to the many lakes, adult bluegill sunfish are not vulnerable visual range of a predator. The mortality risk will thus to predation. This is the situation optimal foraging depend on the vertical attenuation of light. Section theoreticians had in mind (e.g. Schoener, 1987). They 2.3 concerns feeding limitations; feeding rate may be imagined food occurring in food patches, where the limited by handling time, encounter rate and digestion feeders search for these, feed there at maximum.rate, rate. Environmental variability in encounter rate may and then seek shelter while the ingested food is span many orders of magnitude due to differences in digested. Then, for a foraging animal, they assumed light intensity, while there is low vertical variation a proportionality between feeding rate and fitness in digestion rate and handling time. In Section 2.4 we (Emlen, 1966) and developed optimal diet models argue that if the food concentrations are high enough containing the energetic gain from a prey and the to allow for selective feeding at a particular depth in handling time to ingest it. Gilliam's (1990) propor- the water column, there must also exist a depth range tionality between handling time and mortality risk is below this zone where planktivores are encounter- reasonable for this situation. limited. By combining environmental gradients in The pelagic habitat is not like this. It is a three- mortality risk with maximum potential environmental dimensional transparent landscape, without local impact on digestion rate, we finally show in Section shelter adjacent to the feeding areas. True pelagic 3 that if the planktivore enters this depth range where organisms cannot feed efficiently in a food patch and diet selection is possible, it may hardly achieve a then seek cover nearby while digesting. What comes fitness premium. closest to a cover is vertical migration or shoaling. Both these strategies may involve costs and benefits (Clark & Mangel, 1986; Pitcher & Parrish, 1993) and, 2. Models and Data due to their complexity, a thorough discussion is beyond the scope of this paper. 2.1. FITNESSTRANSLATION OF FEEDING AND In this paper, we discuss whether pelagic plankti- MORTALITYRISK vores may have higher reproductive rate by feeding In the original optimal foraging theory, a direct unselectively upon encounter with potential prey than link between feeding rate maximization and fitness by a restricted diet. Our argument is based on the was assumed (Emlen, 1966): by feeding at maximum assumption that the predator of the planktivore will efficiency, time was saved for other fitness-related find its prey by visual stimuli. Although long-distance activities. During the last decade, other aspects of searching must rely on other cues, and other sensory fitness have become explicit, and Mangel & Clark systems will eventually replace vision at large depths (1986) expressed in dynamic programming terms, a OPTIMAL UNSELECTIVE FEEDING 43 framework for which several aspects of fitness could mortality risk from a given day over the rest of the be included simultaneously. By a life-history ap- juvenile period, so that juvenile survival over the proach, direct relationships between behaviour and remaining generation time may be expressed adaptive traits can be revealed. Following Giske et al. S = e -me, we obtain (1993) we express fitness as the difference between p = In b/G - M. (3b) reproductive rate of an individual p and that of its population r Equation (3b) expresses the average importance of fecundity, generation time and mortality for the re- (I) = p -- r. (2) productive rate. Due to the assumption of constant mortality risk, eqn (3b) does not apply to the initial The reproductive rate of an individual can be egg and larval period with high and decreasing mor- expressed as tality rate (Chen& Watanabe, 1989; Roff, 1992: p = ln(bS)/G, (3a) 107-110).